Abstract:Our intuition suggests that when a movie is played in reverse, our perception of motion in the reversed movie will be perfectly inverted compared to the original. This intuition is also reflected in many classical theoretical and practical models of motion detection. However, here we demonstrate that this symmetry of motion perception upon time reversal is often broken in real visual systems. In this work, we designed a set of visual stimuli to investigate how stimulus symmetries affect time reversal symmetry breaking in the fruit fly ’s well-studied optomotor rotation behavior. We discovered a suite of new stimuli with a wide variety of different properties that can lead to broken time reversal symmetries in fly behavioral responses. We then trained neural network models to predict the velocity of scenes with both natural and artificial contrast distributions. Training with naturalistic contrast distributions yielded models that break time reversal symmetry, even when the training data was time reversal symmetric. We show analytically and numerically that the breaking of time reversal symmetry in the model responses can arise from contrast asymmetry in the training data, but can also arise from other features of the contrast distribution. Furthermore, shallower neural network models can exhibit stronger symmetry breaking than deeper ones, suggesting that less flexible neural networks promote some forms of time reversal symmetry breaking. Overall, these results reveal a surprising feature of biological motion detectors and suggest that it could arise from constrained optimization in natural environments.

What problem does this paper attempt to address?

The problem this paper attempts to address is that in visual motion detection, when a video is played in reverse, the perception of motion by organisms is not always exactly opposite to that when played forward, indicating a phenomenon of time-reversal symmetry breaking. Specifically, the authors designed a series of visual stimulus experiments to explore this symmetry breaking in the optomotor response behavior of fruit flies (Drosophila) and analyzed the potential causes of this phenomenon. Through these studies, the authors aim to reveal how biological motion detectors in natural environments produce non-intuitive time-reversal symmetry breaking through optimization processes. Key points of the paper include: 1. **Observed Phenomenon**: Under certain visual stimuli, the behavioral responses of fruit flies exhibit time-reversal symmetry breaking, meaning that their motion perception does not simply reverse when the video is played backward. 2. **Experimental Methods**: Various visual stimuli with different characteristics were designed to test the responses of fruit flies under different conditions, including stationary sawtooth contrast and local light edges. 3. **Theoretical Model**: Neural network models were trained to predict the speed of scenes. These models were optimized based on the contrast distribution of natural scenes. The results showed that even if the training data itself is symmetrical, the model's response can still break time-reversal symmetry. 4. **Conclusion**: Time-reversal symmetry breaking is related not only to the characteristics of the stimuli but also to the architecture and optimization of the detectors. Particularly when processing natural inputs, the optimization process may lead to symmetry breaking, which could be a way for biological systems to adapt to natural environments. Through these studies, the paper provides new perspectives on understanding symmetry and asymmetry in biological visual systems, contributing to a deeper understanding of biological motion detection mechanisms and their adaptability in natural environments.

Broken time reversal symmetry in visual motion detection

Learning Broken Symmetries with Resimulation and Encouraged Invariance

Natural Scene Statistics and the Structure of Orientation Maps in the Visual Cortex

Quantifying stimulus-relevant representational drift using cross-modality contrastive learning

A Rotational Motion Perception Neural Network Based on Asymmetric Spatiotemporal Visual Information Processing

Neural mechanisms to incorporate visual counterevidence in self-movement estimation

Retinal waves prime visual motion detection by simulating future optic flow

Pre-training artificial neural networks with spontaneous retinal activity improves motion prediction in natural scenes

Asymmetry in Visual Cortical Circuits Underlying Motion-Induced Perceptual Mislocalization

What happens if it changes color when it moves?: psychophysical experiments on the nature of chromatic input to motion detectors

Modeling bio-inspired visual neural for detecting visual features of small- and wide-field moving targets synchronously from complex dynamic environments

Adaptation of retinal discriminability to natural scenes

Rigid Propagation of Visual Motion in the Insect’s Neural System

Modeling Drosophila vision neural pathways to detect weak moving targets from cluttered backgrounds

Divergent mechanisms of perceptual reversals in spinning and wobbling structure-from-motion stimuli

Corrective mechanisms of motion extrapolation

Learning Broken Symmetries with Approximate Invariance

Modeling spatial contrast sensitivity in responses of primate retinal ganglion cells to natural movies

Artificial Visual Network with Fully Modeled Retinal Direction-Selective Neural Pathway for Motion Direction Detection in Grayscale Scenes

Probing the effects of broken symmetries in machine learning

Perception of opposite-direction motion in random dot kinematograms